You can upload existing data to your project in the
Data Acquisition Format (CBOR, JSON, CSV),
or as WAV, JPG, PNG, AVI or MP4 files.
We also support uploading image datasets with labels in various formats. When you include labels during upload,
we attempt to convert your dataset into a format recognized by Studio.
Learn more here.
Bounding boxes: You can upload object detection datasets in
any supported format.
Select both your images and the label file(s) when uploading to apply the labels.
Using the info.labels file for labels and category.
Image label format
Annotations in this format could not be found in the selected files.
Select both your images and any label files when uploading to apply the labels.
A label map file could not be detected.
This format requires a label map file, which maps keys to the label they represent.
You can fix these labels later by clicking 'Edit labels' on the data acquisition page.
Upload into category
Upload category will be derived from the structure of your dataset
(e.g. samples in a 'train' directory will be uploaded into training data).
You need to specify a label
This dataset format uses bounding box labeling, used for object detection.
The project labeling method will switch to 'bounding boxes'.
This dataset format uses one label per sample. You may wish to change your project
labeling method to 'one label per data item' in the project dashboard.
The selected samples contain the following labels. Which ones do you want to edit?
Set the label
Dataset train / test split ratio
Training data is used to train your model, and testing data is used to test your model's accuracy after training.
We recommend an approximate 80/20 train/test split ratio for your data for every class (or label) in your dataset, although especially large datasets may require less testing data.
Suggested train / test split
80% / 20%
Labels in your dataset
Perform train / test split
Use this option to rebalance your data, automatically splitting items between training and testing datasets. Warning: this action cannot be undone.